01
What actually drives the model?
The workbook contains formulas and dependencies. IRIS turns that machinery into an intelligible account of the economic system.
We build the context layer between domain expertise and frontier intelligence.
Our first product is IRIS, the AI harness for financial analysts.
The sense-making overlay for consequential analytical models. Import a model: IRIS creates a plain-English How this model works, identifies the key forecast judgments and their reach, and makes each judgment an English-readable, LLM-runnable Method. Every Method, model run, and result is versioned and snapshotted, so the work can be understood, interrogated, changed, and remembered.
The premise
The workbook contains the model. Not its durable explanation.
Spreadsheets are good at preserving calculations. Frontier models are good at reasoning. Neither, by itself, preserves the meaning an analyst needs to understand and change a consequential model.
02
Which variables are judgments versus consequences?
IRIS separates uncertain inputs the analyst must forecast from values determined by the model's logic.
03
Why does the current forecast look this way?
Methods connect each important forecast to the reasoning, assumptions, and evidence behind it.
04
What should change the view?
The relevant evidence and known unknowns stay attached to the judgment they inform.
05
What does a changed view do to the model?
Execution turns revised judgment back into numbers and carries the consequence through the workbook.
06
How did the thinking evolve?
Versions, Runs, Results, and Timeline preserve what changed, why it changed, and what happened next.
IRIS
Make the model legible.
IRIS connects model truth, economic meaning, analyst judgment, execution, and history in one durable system.
The architecture
Import the model. Understand the system. Drill into the judgments.
IRIS turns an unfamiliar workbook into a model-level explanation and a map of the judgments inside it. From there, the analyst can open any judgment as a Method—without losing the cells, formulas, evidence, or history underneath it.
01 / IMPORT
Bring the model
Import the workbook the analyst already uses. Excel remains the model; IRIS begins with its actual cells and formulas.
02 / EXPLAIN
How this model works
Deterministically establish periods, formulas, dependencies, assumptions, historical series, and current forecasts—then explain the system in plain English.
03 / JUDGMENTS
Drill into the judgments
Identify the variables an analyst must judge, show which reach the most of the model, and separate them from calculated consequences.
04 / METHOD
Read and edit the Method
Open any judgment to see its current forecast and model-derived reasoning: how to forecast it, what evidence matters, and when the view should change.
05 / RUN
Put judgment back into numbers
Execute the Method through =IRIS.AI(), preview the result, and carry the changed view through the workbook.
06 / SNAPSHOT
Version everything
Snapshot the model, Method, evidence, Run, and Result. Preserve what changed, why it changed, and what it affected.
The model becomes working context for people and AI—not just a collection of cells and formulas.
Understanding and judgment
Know how the model works. Know where judgment enters.
IRIS separates the model's deterministic structure from the uncertain variables an analyst must actually forecast.
How this model works explains the system.
IRIS reads formulas, values, dependencies, and historical relationships, then compresses them into an economic explanation of the model's structure and drivers. It narrows 11,112 cells to the question that matters without losing the model around it.
Methods explain how uncertain variables should be forecast.
Each judgment has an identity, Method, evidence, dependencies, and known unknowns. The analyst can edit the Method in plain English, execute it through =IRIS.AI(), and inspect the consequence in the model.
The product
Understand. Interrogate. Change.
The model becomes a legible, interactive system rather than a grid that must be reverse-engineered every time.
01 / Interrogate
Ask the model the question you actually have.
Select an output, assumption, or driver. Ask why it changed, what drives it, or what the analyst believed last time.
02 / Understand
Get the model in plain English.
IRIS connects the formulas to the Method, assumptions, evidence, and history behind the answer.
03 / Change
Change the view and see the consequence.
Edit the judgment in plain English, run it in the model, and trace the result back to the Method, evidence, Version, and prior call.
No more model archaeology.
Stop tracing precedents and reconstructing someone else's logic before you can answer a basic question.
A new, but familiar, way of working.
IRIS owns the analytical model and its history. Excel is an analyst-native client of it.
Model memory
Remember how the thinking changed.
IRIS preserves the prior call, the facts behind it, what changed, why it changed, and what the judgment affected.
Every answer and model change can be traced to the exact company, model, subject, Method, Version, evidence, and lineage that produced it.
Exact identities. Traceable evidence. Versioned judgment. Durable analytical memory.
The payoff
Make the model understandable—to people and AI.
IRIS is the sense-making layer for consequential analytical models.